@ Griffith: Can you explain a little more about the "stepping" in your last post? Thanks.
@ CoolColJ: No arguing that Shift didn't get the sound right. Maybe I've enjoyed the game so much I haven't studied the sounds the way you and Griffith have. You guys seem to be more sensitive to sound issues than my laymen's ears allow me to be. Still, won't you admit it's easier to get excited over the raw (if overdone) sounds in Shift versus GT?
I will agree that I have spent a lot of time studying sounds, largely as a personal project over the last few years, but I've also been playing games for most of my life (well, cracking on for 80%+ anyway) and most of that time has been spent on racing games.
Basically, I know what I hear,
but I know that there's a lot my brain will make me
think I can hear, so I try hard to listen through that. But anyway, that's another discussion that I've bored people with in the past
As for the stepping, is quite tricky to explain. Basically, it sounds to me like the parameters used to govern the playback of the samples being used as engine sounds is not updating "fast enough". What is fast enough? Well at least 36 Hz (or some higher multiple of it, going by Grand Prix Legends, which still isn't
quite fast enough for my ears.)
All that needs to be updated to change the sound is throttle position and rpm - two numbers, per car. Given that physics calculations need tens of numbers, probably more, to describe the status of a car, adding two more at the physics rate is no big deal. Typical physics rates these days are in excess of 100 Hz, which ought to be fine for my ears - 0.01 seconds is a normal figure for the "discrete sound" threshold our brains have - any two sounds with a gap of less than this threshold sounds like one complete sound to us, in other words.
I just tested myself to verify this, and it's important to note that it depends highly on the manner / nature of the sound and the listening environment: using headphones and a synthetic AK47 sound (without reverb), I could discern two separate "bangs" down to 6 milliseconds, or 167 Hz. I "knew" that the sound I was hearing was two separate events, so even though it began overlapping at around 11 milliseconds (91 Hz), I could still "tell" they were separate - if I'd done it blind, or with a strong reverb, it may have been a different story - i.e. a lower (Hz) "score".
------------------
Anyway, the graph in the "Modulation" section of
this article shows what I mean by stepping; it's shown below for clarity. The governing data (i.e. rpm, throttle position) are sampled / interpolated at a given rate and this gives the "stepped" approximation (in black) of the "real" status (red line) of the data (really only as fine / accurate as the time-step of the game engine itself, in this case). If this update rate is less than a given value (e.g. 100 Hz, as above) then the stepping may be audible, as indeed it is in many, many racing games - particularly when the signal changes quickly: revving, gear changes (1st-2nd particularly) and doppler shift as the cars pass
close to the camera, maximising the error over time (the space between the red and black lines).
It's important to remember that the above does not represent the actual sound / it's waveform, rather the control signals only; but the sound is dependent on the control signals, so any sudden steps in the control will give a corresponding step in the resultant sound, according to the severity of the control step and the character of the sound it governs.
When you also realise that audio is typically
processed at around 44
kHz, you can see why anything less than 100 Hz just seems, well,
stupid, because the audio engine has thrown out at least 441 samples before the rpm or throttle position has been "updated".
If you need me to, I can throw together a demo of how sampling rate of "control data" affects the sound of some poorly synthesised engine "noises"
(Sorry for the large post!)